{"id":"W2969250926","doi":"10.1080/0142159x.2019.1652260","title":"Are rating scales really better than checklists for measuring increasing levels of expertise?","year":2019,"lang":"en","type":"article","venue":"Medical Teacher","topic":"Clinical Reasoning and Diagnostic Skills","field":"Medicine","cited_by":20,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"University of Ottawa","keywords":"Checklist; Rating scale; Psychology; Test (biology); Measure (data warehouse); Objective structured clinical examination; Medical education; Rating system; Applied psychology; Medicine; Clinical psychology; Computer science; Data mining; Cognitive psychology; Developmental psychology","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1353231,0.001175156,0.002229912,0.004636266,0.0006858978,0.00361156,0.002085524,0.003128338,0.003639417],"category_scores_gemma":[0.4546559,0.0007710999,0.001531454,0.006226876,0.004877732,0.009803807,0.001351767,0.002232764,0.00135448],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001326756,"about_ca_system_score_gemma":0.001197061,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002145814,"about_ca_topic_score_gemma":0.003216412,"domain_scores_codex":[0.8530302,0.09375188,0.01512351,0.008226496,0.02843438,0.001433507],"domain_scores_gemma":[0.3734806,0.4591019,0.07890399,0.03232223,0.0525094,0.003681891],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.002119409,0.000480148,0.6313459,0.003502915,0.003149495,0.000172083,0.003207124,0.002288743,0.002246825,0.006615188,0.01331657,0.3315556],"study_design_scores_gemma":[0.0003122011,0.004273424,0.9318342,0.004966339,0.0009576115,0.001110393,0.003242001,0.01044458,0.003615434,0.0170488,0.02174857,0.0004465157],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6430911,0.05458119,0.2076319,0.03495865,0.008091587,0.001183694,0.003035922,0.002057735,0.04536823],"genre_scores_gemma":[0.9101883,0.004688986,0.07790265,0.003366555,0.001358458,0.0006078749,0.0005778408,0.0002637245,0.001045652],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.864677,"threshold_uncertainty_score":0.7156653,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06439051680890505,"score_gpt":0.3509525032084154,"score_spread":0.2865619863995103,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}